INDEX

Explanations

assistant responses

np_acts-logits-general · gemini-2.5-flash-lite

common function words—especially prepositions and conjunctions (even across languages)—and sometimes structural header markers.

oai_token-act-pair · gpt-5 Triggered by @vetterc0

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.3-70B-IT @ 50-resid-post-gf

Configuration

Goodfire/Llama-3.3-70B-Instruct-SAE-l50/Llama-3.3-70B-Instruct-SAE-l50.pt

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 aforementioned

-0.07

(?

-0.07

(?)

-0.07

(>

-0.07

,,

-0.07

is

-0.07

en

-0.07

(the

-0.06

-\n

-0.06

POSITIVE LOGITS

¶Į

0.15

 -*-č\n

0.13

ÐĲÑĢÑħÑĸÐ²

0.12

<|begin_of_text|>

0.11

******č\n

0.11

.Formatter

0.10

įng

0.10

Ð©Ð¾

0.10

¦æĥħ

0.10

printStats

0.10

Activations Density 1.375%

assistant responses

common function words—especially prepositions and conjunctions (even across languages)—and sometimes structural header markers.

No Comments

No Known Activations

assistant responses

common function words—especially prepositions and conjunctions (even across languages)—and sometimes structural header markers.

No Comments

No Known Activations